基于智能代理的交通分配建模

Traffic assignment model based on intelligent agent

针对现有交通分配研究对驾驶员有限理性、道路交通动态随机性的忽视,提出一种基于智能代理的动态交通分配模型IATAM,其以驾驶员为代理,依据驾驶员的路线偏好,综合驾驶员信息处理的模糊随机过程,考虑邻居驾驶员影响,提出神经网络结构的驾驶员智能路线决策机制。实验中IATAM模型检测点车流量与真实车流量的平均相对误差减小到6.42%,表明IATAM模型实验精度更高,更符合复杂多变的交通环境。基于智能代理进行交通分配,更能反映驾驶员的异构性和路线决策的模糊随机性,并有效提高交通分配精度。

On account of the neglect of drivers＇ bounded rationality and traffic dynamic and random in the existing traffic assignment study,the paper proposes the traffic assignment model based on intelligent agent（ IATAM）. Each driver is defined as an agent and has different route preference. Based on the composition of fuzzy stochastic information processes of all drivers,as well as the effect of neighbor drivers,the agent route decision-making mechanism is constructed which is like neural network. Average Relative Error（ ARE） of IATAM model and actual road network in checkpoint volume is reduced to 6. 42%,indicating that the IATAM model has higher prediction precision and conforms more to the complex traffic environment. Traffic assignment based on intelligent agent reflects the isomerism of drivers and the vagueness and randomness of route decision-making,which improves the accuracy of traffic volume significantly.